OpenVIDIA : Parallel GPU Computer Vision

Tracker

 - Bugs ( 0 open / 0 total )
Bug Tracking System

 - Support Requests ( 1 open / 1 total )
Tech Support Tracking System

 - Patches ( 1 open / 1 total )
Patch Tracking System

 - Feature Requests ( 0 open / 1 total )
Feature Request Tracking System

 Forums

 Doc Manager

 Mailing Lists

 Screenshots

 Task Manager

 CVS Tree

 Released Files

Related Projects

Comparametric Toolkit

Reference Links

Open GL Reference

nVIDIA cg homepage

GeForce FX Overclocking

GPGPU

gpu's arent just good for vision and computation - some clever folks have started using them for games as well. ;) har har linux-games.net

7 Years of Graphics Card History

Page design based on Blosxom (which was used originally before the move to sourceforge).

This page best viewed with Dillo, Lynx, w3m, Mozilla-Firefox, Galeon, Epiphany, konqueror . . .

Screenshots & Videos

Parallel Graphics Cards A Cool, but somewhat dated machine used by OpenVIDIA to explore parallelism. This photograph of the machine shows 6 PCI GeForce FX 5200 cards and one AGP GeForce card.
glestpchirp2m VideoOrbits image homography running on the University College sequence	Canny edge filter (implemented by Fung and Aimone)
Modified Canny edge filter: Locates corners instead (Fung and Aimone) OpenVIDIA locates corners and edges and calculates feature vectors for each corner in realtime on a FX6800 Click the image for a movie. [mpg 1.2 MB]	All the world's a marker: Markerless tracking The above altered to track corner points in the image (Fung, Aimone). The green lines connect the same point in the previous image to its new location in the next image. Click the image for a movie. [mpg 6.0 MB high quality to see the lines properlly]
Download Video Corner tracking + Projective RANSAC motion estimations. Blue dots indicate found corners, green lines connect the points to its previous location, indicating corners which all have the same, dominant motion. Notice that the moving hand and unmatchable corners are segmented out, and do not affect the estimation on the main projective motion. It can track over 140 correspondences at 30 fps. This video is live (you can see it when it looks at the monitor too), but is running at 22 fps on account of having to compress and write a jpeg to disk every frame (in order to make the video).	Features are tracked and given an ID number Possibly useful input for higher level vision tasks. [16MB video].
Skin tone segmentations: Locates people and hands, and is a useful first step in many applications, including gesture tracking. The resulting load on a 2.0 GHz CPU machine was 0.02, 30fps, 320x240
	3D registration. Imported objects from Cal3D (import conducted thanks to Mohit Kansai and Billal Belmellat, 3d registration thanks to Chris Aimone).
flcornersTrack scene tracking/annotation program. More info Click for a video (1MB).	3D studio import. [v0.07]